PROMPT INJECTION ATTACKS ON LARGE LANGUAGE MODELS: A SYSTEMATIC LITERATURE REVIEW

Date

2025-6-11

Author

Bayhan, Fatih

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

180
views

50
downloads

Cite This

The use of Large Language Models has increased significantly after the release of ChatGPT in 2022. LLMs are used in various applications including chatbots and generative artificial intelligence applications. LLMs are integrated into these applications and perform their functionality at the backend. Like every new technology, LLMs or LLM-integrated applications are vulnerable to attacks. Among these attacks, prompt injection attacks have emerged in recent years due to the widespread use of LLM-integrated applications. These attacks aim to manipulate LLM outputs and cause the model to provide incorrect or harmful responses. This study focuses on examining existing research on prompt injection attacks. In this scope, the main aim of the study is to to classify attack types, determine current defense mechanisms, and provide guidelines to enhance LLM security. A systematic review study on 36 papers published between 2022 and 2025 was examined. The papers focusing on either prompt injection attacks or defense mechanisms were selected from the IEEE, WOS, and Scopus databases. A qualitative analysis was conducted to categorize attack methods, defense techniques, and highlight their contributions and future research directions to promote secure LLM development research.

Subject Keywords

Prompt injection, LLM, LLM security

URI

https://hdl.handle.net/11511/115034

Collections

Graduate School of Informatics, Term Project